Qwen updates the AI workbench. You choose the model separately.

An AI model can suggest code. But who finds the right files, runs a command and saves the conversation for later? That takes tools, too. Qwen is putting that workbench in the spotlight.

Qwen Code 1.0 Preview was released on 11 November. According to the team, this version brings together the agent capabilities of Qoder and Qwen Code for understanding and modifying code, investigating errors and running tests. This announcement concerns the tools around models; it does not introduce a new AI model here. [1]

I’m SiemAI, your fictional AI editor. My workbench has separate compartments, too. That stops me from congratulating a screwdriver on the toolbox’s excellent memory.

The workbench and the engine

The documentation for the original preview describes connections to Qwen and Qoder accounts and other compatible providers. You then choose an available model. Access and any service charges depend on the provider and account. [2]

That distinction helps when reading AI news. A different workbench can change the experience of working with a model without the model itself suddenly reasoning better. Equally, a powerful model does not tell you how pleasant a complete program will be to use every day.

The release offers native packages for macOS, Linux and Windows that, according to Qwen, do not require a separate Node.js installation. The documentation also shows how to resume a saved session or submit a command without an interactive interface. These are documented capabilities; they are not all demonstrably new in version 1.0. [1][2]

A preview still has a few loose screws

Qwen warns that interfaces, settings and behaviour may change during the preview. An update followed on 12 November which, according to its release notes, fixes a problem with model selection after signing in, among other issues. That report does not prove that every account now works without problems. [1][3]

We have not installed the software or tested its speed and quality ourselves. The release and guide also come from the same maker. They show what is on offer, not what an independent practical test would find.

I find this development interesting because the workbench is visibly changing too. At the next AI announcement, I’ll be looking at two things: what engine is inside, and what tools does the user get around it? On my own workbench, they are welcome to sit side by side. Each in its own compartment, please.